1 Lysosomal membrane bind cholesterol and

2 contribute to lysosomal cholesterol export

3

4 Jian Li and Suzanne R. Pfeffer*

5 Department of Biochemistry, Stanford University School of Medicine 6 Stanford, CA USA 94305-5307 7

8 *Correspondence to: [email protected].

9

10 Abstract: LAMP1 and LAMP2 are highly abundant, ubiquitous, mammalian proteins

11 that line the limiting membrane, and protect it from lysosomal hydrolase action.

12 LAMP2 deficiency causes Danon’s disease, an X-linked hypertrophic cardiomyopathy. LAMP2

13 is needed for chaperone-mediated autophagy, and its expression improves tissue function in

14 models of aging. We show here that human LAMP1 and LAMP2 bind cholesterol in a manner

15 that buries the cholesterol 3β-hydroxyl group; they also bind tightly to NPC1 and NPC2 proteins

16 that export cholesterol from . Quantitation of cellular LAMP2 and NPC1 levels

17 suggest that LAMP proteins represent a significant cholesterol binding site at the lysosome

18 limiting membrane, and may signal cholesterol availability. Functional rescue experiments show

19 that the ability of human LAMP2 to facilitate cholesterol export from lysosomes relies on its

20 ability to bind cholesterol directly.

21

22 23 Introduction

24 Eukaryotic lysosomes are acidic, membrane-bound organelles that contain , lipases and

25 nucleases and degrade cellular components to regenerate catabolic precursors for cellular use (1-

26 3). Lysosomes are crucial for the degradation of substrates from the , as well as

27 membrane bound compartments derived from the secretory, endocytic, autophagic and

28 phagocytic pathways. The limiting membrane of lysosomes is lined with so-called lysosomal

29 membrane glycoproteins (LAMPs) that are comprised of a short cytoplasmic domain, a single

30 transmembrane span, and a highly, N- and O-glycosylated lumenal domain (4-6). Because of

31 their abundance and content, LAMPs have been proposed to serve as a protective barrier

32 to block hydrolase access to the limiting phospholipid bilayer. LAMP1 and LAMP2 are 37%

33 identical and may overlap in function, but knockout of LAMP1 in mouse has a much milder

34 than depletion of LAMP2 (7): LAMP2-deficient mice have very short lifespans, and

35 show massive accumulation of autophagic structures in most tissues. Indeed, LAMPs are

36 required for fusion of lysosomes with phagosomes (8) and LAMP2 has also been proposed to

37 serve as a receptor for chaperone-mediated autophagy (9-11).

38

39 Previous work has implicated LAMP2 in cholesterol export from lysosomes, as LAMP-deficient

40 cells show cholesterol accumulation that can be rescued by LAMP2 expression (12,13).

41 Proteome-wide analysis of cholesterol binding proteins included LAMP1 and LAMP2 among a

42 long list of candidate proteins (14). Despite these hints, the precise function of LAMP proteins

43 has remained unclear, and they are often presumed to be structural components. We show here

44 that LAMP proteins bind cholesterol directly and this capacity contributes to their role in

45 cholesterol export from lysosomes.

2 46 Results and Discussion

47 We sought to verify direct cholesterol binding to LAMP proteins using LAMP protein lumenal

48 domains, engineered to be secreted from cells by simple deletion of their transmembrane and

49 short cytoplasmic domains (Fig. 1-figure supplement 1; Fig. 1A). Soluble, purified, LAMP1 and

50 LAMP2 proteins appeared to bind 3H-cholesterol saturably, at a stoichiometry comparable to

51 equimolar amounts of purified, NPC1 N-terminal domain (NTD) that contains a single

52 cholesterol binding site (15,16; Figs. 1B, D, E). (Note that this does not provide information

53 about relative binding affinities.) Binding was not especially sensitive to pH (Fig. 1H) and was

54 complete after ~2 hours at 4°C (Fig. 1I).

55

56 Cholesterol is poorly soluble, thus binding reactions were carried out in the presence of sub-

57 critical micelle concentration amounts of Nonidet P40 detergent (0.004%) to help solubilize the

58 cholesterol, as worked out by Infante et al. in their studies of cholesterol binding to the N-

59 terminal domain of NPC1 protein (15). Under these conditions, most of the cholesterol remains

60 in a mixed micelle of cholesterol and detergent and is still poorly soluble. Thus, the apparent

61 affinity for cholesterol is likely to be tighter than the curves indicate, as the amounts added do

62 not reflect the concentration of free cholesterol that is actually available for binding.

63

64 Preliminary experiments showed that 3H-cholesterol binding was competed by unlabeled

65 cholesterol, 24-hydroxycholesterol, but not cholesterol sulfate (Fig. 1G). This suggested that

66 binding occurs via the 3β-hydroxyl moiety of the cholesterol molecule, similar to the orientation

67 with which the NPC1 N-terminal domain binds cholesterol (16). Consistent with this, LAMP2

68 also bound 3H-25-hydroxycholesterol with similar apparent affinity as cholesterol; binding was

3 69 competed by excess cold 25-hydroxycholesterol (Fig. 1C), as would be expected for a specific

70 interaction. Only low levels of background binding were detected using GFP-binding protein or

71 TIP47 as controls (Fig. 1 D,E). More detailed analysis confirmed that 25-hydroxycholesterol

72 (Fig. 2B) and 7-ketocholesterol (Fig. 2D), but not cholesterol sulfate (Fig. 2A), compete with 3H-

73 cholesterol for binding to LAMP2 protein.

74

75 Epicholesterol is a cholesterol epimer that differs only in the chirality of carbon 3 such that the

76 hydroxyl group is in the alpha rather than beta conformation. Importantly, epicholesterol failed

77 to compete for cholesterol binding to LAMP2 protein under conditions where cholesterol

78 competed for binding (Fig. 2C). It was not possible to add higher concentrations of sterol

79 competitors due to solubility issues, but significant inhibition was observed. Together, these data

80 strongly support the conclusion that cholesterol binds LAMP2 via its 3β-hydroxyl moiety.

81

82 A slight stimulation of binding was seen in reactions containing low levels of competitor

83 cholesterol sulfate or 25-hydroxycholesterol (Fig. 2A,B); this is presumably due to the higher

84 solubility of these sterols, which will help solubilize 3H-cholesterol present in the reaction’s

85 mixed micelles, and presumably make it more available for LAMP2 binding (see also ref. 15).

86 Despite its somewhat higher solubility, 25-hydroxycholesterol did not appear to bind LAMP2

87 much more tightly than cholesterol, at least as inferred from its ability to compete with

88 cholesterol for binding (Figure 2B) or to bind directly (Fig. 1C).

89

90 The LAMP protein family includes LAMP1, LAMP2, DC-LAMP, BAD-LAMP and Macrosialin

91 (4). Each of these proteins contains a related “LAMP” domain; LAMP1 and LAMP2 proteins

4 92 each contain two (Fig. 1-figure supplement 1). Soluble versions of the individual, membrane

93 distal (“domain 1”) and membrane proximal (“domain 2”) LAMP domains of LAMP2 (Fig. 1A)

94 bound cholesterol with different capacity: the N-terminal, domain 1 bound more cholesterol than

95 its membrane proximal, domain 2 counterpart under these conditions (Fig. 1F). It is possible that

96 both are capable of binding cholesterol within the full length molecule, as total domain 1 binding

97 was less than that seen with the full length, secreted LAMP2 construct (Fig. 1F).

98

99 To verify that LAMP2 binds cholesterol in cells, soluble LAMP2 protein was expressed and

100 purified from the secretions of HEK293F cells grown in protein-free, FreeStyle 293 Expression

101 Medium that does not contain cholesterol. Under these conditions, any LAMP2-bound sterol

102 must come from intracellular sources. We subjected freshly purified LAMP2 protein to

103 chloroform:methanol extraction and analyzed the extract by thin layer chromatography. As

104 shown in Fig. 3C, the LAMP2 extract contained a molecular species that co-chromatographed

105 with cholesterol but not 24-hydroxy-, 25-hydroxy-, or 26-hydroxycholesterol, lanosterol or 7-

106 beta-hydroxycholesterol. Mass spectrometry of the eluted material (Fig. 3B) confirmed a profile

107 identical with purified cholesterol standard (Fig. 3C). These experiments show that LAMP2

108 purified from cell secretions carries primarily, bound cholesterol.

109

110 NPC1 and NPC2 proteins mediate cholesterol export from lysosomes (16,17). NPC1 has 13

111 transmembrane domains, and three large, lumenal domains that are important for its function.

112 As mentioned earlier, the NPC1 N-terminal domain binds cholesterol directly (16). Because of a

113 possible connection between LAMP protein cholesterol binding and NPC-mediated cholesterol

114 export, we checked for an interaction between these proteins. Membrane anchored, endogenous

5 115 LAMP2 co-immuno-precipitated with full length NPC1-GFP but not with the control protein,

116 GFP (Fig. 4A) or the lysosomal membrane protein, MCOLN1 (Fig. 4D), upon expression in

117 HEK293T cells (Fig. 4A). Interestingly, co-immunoprecipitation decreased in cells treated for 24

118 hours with cyclodextrin to remove cholesterol from lysosomes (18,19) and the plasma membrane

119 (Fig. 4B, C). These conditions (~ 0.1% cyclodextrin) have been shown to be non-toxic (cf. 20)

120 and did not alter cell growth rate or viability in our hands.

121

122 Purified, soluble LAMP2 protein also bound very tightly (and directly) to the N-terminal domain

123 of NPC1 protein (Fig. 5A, red) in the presence (or absence, not shown) of cholesterol (KD=6nM),

124 as monitored by microscale thermophoresis using AF647 dye-conjugated NPC1 protein—

125 binding significantly altered the fluorescence of NPC1 protein (Fig. 5C). No interaction was

126 observed for the control , RNase B (Fig. 5E). The smaller, NPC2 protein (Fig. 5B)

127 also bound to LAMP2 (KD=122nM), but ~20 fold less tightly that NPC1 N-terminal domain

128 (Fig. 5D); binding was monitored in the presence of cholesterol sulfate which will occupy the

129 binding site of NPC2 but not LAMP2 (Figs. 1 and 2). No binding was seen for NPC2 to the

130 control RNase B protein (Fig. 5F). Thus, LAMP2 binds directly to both NPC1 N-terminal

131 domain and NPC2 proteins. For NPC1, the enhanced binding seen in cells in the presence of

132 cholesterol does not appear to reflect occupancy of NPC1’s N-terminal domain, as this variable

133 did not influence LAMP2 binding in solution. However, it is important to note that NPC1 also

134 likely binds cholesterol within its membrane spanning region which may also influence its

135 overall conformation (21,22).

136

6 137 LAMP proteins are highly abundant components of the lysosome membrane and may serve as a

138 reservoir for cholesterol extracted from intra-lysosomal membranes by NPC2, prior to

139 cholesterol export from lysosomes by NPC1. [The term, reservoir, is meant to imply a holding

140 station for cholesterol molecules that have been solubilized from the internal lipid contents of

141 lysosomes by NPC2, and held closer to the limiting membrane, prior to NPC1-mediated export.]

142

143 Are LAMP proteins abundant enough to represent a cholesterol reservoir? We used purified

144 LAMP2 and NPC1 proteins as standards to determine their precise abundance in HeLa and

145 HEK293 cell lysates (Fig. 5-Fig. supplement 1). Using the polypeptide molecular weights and

146 cellular protein determinations, we estimate that HeLa cells contain 6.8 X 106 LAMP2 molecules

147 per cell and 3.7 X 105 NPC1 molecules per cell, or 18 fold more LAMP2 than NPC1; HEK293T

148 cells contain 2 X 106 LAMP2 molecules per cell and 5.9 X 105 NPC1 molecules (3.6X fold more

149 LAMP2).

150

151 Baby hamster kidney cells have been estimated to contain an absolute volume of ~37µm3

152 lysosomes and prelysosomes per cell (3.7 X 10-14 l) and a lysosome membrane area of 370µm2

153 (23). Assuming similar values for HeLa cells, this would represent a LAMP2 membrane density

154 of 18,378 or 5676 molecules per µm2 in HeLa or 293T cell lysosomes, respectively, consistent

155 with previous reports (6). For comparison, tightly packed viral spike glycoproteins occur at a

156 density of 22,000 molecules per µm2 (24). We assume that LAMP1 will be of similar high

157 density, together with LAMP2, practically lining the interior of the lysosome limiting membrane.

158 In terms of concentrations, 37 femtoliters of lysosome volume would contain 0.3mM LAMP2-

159 associated cholesterol binding sites in HeLa cell lysosomes (assuming one mole cholesterol

7 160 bound per more LAMP2). It does not seem unreasonable to consider this as a significant

161 reservoir of cholesterol molecules that may be poised for transfer to NPC1 protein prior to

162 export.

163

164 Residues needed for cholesterol binding are needed for LAMP protein function

165 The structure of an individual LAMP domain from DC-LAMP protein is comprised of a novel,

166 beta-prism fold that appears to contain a hydrophobic pocket (4); we used this structure to model

167 the structure of LAMP2 domain 1 (Fig. 6A). Site directed mutagenesis of hydrophobic residues

168 predicted to line the walls of this cavity yielded purified LAMP2 proteins with impaired

169 cholesterol binding activity. Thus, a soluble, LAMP2 domain 1-I111A/V114A construct yielded a

170 secreted protein (Fig. 6B inset, right lane) that bound significantly less cholesterol than its wild

171 type counterpart (Fig. 6B inset, left lane and panel B). Because these proteins were obtained

172 from cell secretions, they are likely to be properly folded, as they escaped the endoplasmic

173 reticulum’s quality control machinery. These experiments show that residues facing the

174 predicted, prism fold pocket are important for cholesterol binding and likely contribute to the

175 cholesterol binding site.

176

177 Finally, to verify the importance of cholesterol binding to LAMP2 protein as part of its

178 physiological role, we tested the ability of wild type and mutant LAMP2 constructs to rescue the

179 cholesterol accumulation seen in lysosomes from mouse embryonic fibroblasts missing LAMP1

180 and LAMP2 proteins (12,13). The ability of lysosomes to export cholesterol can be monitored

181 by feeding cells cholesterol in the form of LDL, and using conversion of 14C-oleic acid to

182 cholesteryl oleate that takes place after endocytosed cholesterol is transported to the endoplasmic

8 183 reticulum (25). Previous work showed that LAMP1/2 knockout MEF cells were impaired in

184 cholesterol export using this assay (13).

185

186 We used lentivirus transduction to test the ability of full length LAMP2, a membrane anchored

187 LAMP2 domain 1 (LAMP2-GFP Δ194-368; Fig. 1-figure supplement 1), or a membrane

188 anchored LAMP2 domain 1-I111A/V114A to rescue the ability of LAMP1/LAMP2 knockout MEF

189 cells to export LDL-derived cholesterol from lysosomes. For these experiments, we used LAMP

190 constructs containing a single LAMP domain, as full length LAMP2 constructs with mutations in

191 both LAMP domains failed to fold properly or be transported efficiently to lysosomes.

192

193 It was important to first verify the precise amounts of each construct in lysosomes, to evaluate

194 any functional rescue findings. Flow cytometry analysis showed that the rescue constructs were

195 expressed at comparable levels in each stably expressing cell population (Fig. 6C). Light

196 microscopy confirmed that the constructs were capable of proper lysosome localization, as

197 determined by their colocalization with endogenous LAMP1 protein (Fig. 6E) in HeLa cells.

198 (Similar staining was observed in LAMP knockout MEF cells that lack LAMP protein markers).

199

200 To fully confirm the folding of these artificial constructs, we analyzed their status

201 and stability after addition of cycloheximide to inhibit new protein synthesis (Fig. 6—Figure

202 supplement 1). Full length GFP-LAMP2 protein migrated at ~140kD and its abundance was not

203 altered after 4 hours cycloheximide treatment, consistent with its long half life in cultured cells

204 (panel B). Similarly, the GFP-domain 1 construct was stable under these conditions and

205 migrated at ~90kD (panels A,B). In contrast, the I111A/V114A mutant domain I protein displayed

9 206 two distinct bands; the upper band was stable, while the lower band likely corresponded to an ER

207 form that was largely degraded after 4 hours in cycloheximide (panels A,B). From this we

208 conclude that cells expressing membrane anchored LAMP2 domain 1 I111A/V114A are less

209 efficient at folding the protein but some folded protein makes it to lysosomes, where it is stable.

210 This difference was accounted for in subsequent functional rescue experiments (Fig. 6D).

211

212 Figure 6D shows that as expected, full length, wild type LAMP2 rescued cholesterol export in

213 LAMP1/2-deficient MEF cells; membrane anchored LAMP2 domain 1 showed a level of rescue

214 consistent with its lower capacity for cholesterol binding (cf. Fig. 1F). Importantly, membrane

215 anchored, LAMP2 domain 1 I111A/V114A failed to rescue cholesterol export from lysosomes

216 (Fig. 6D), consistent with its inability to bind cholesterol; shown are the data corrected for the

217 amount of mature proteins present in lysosomes in these cells. LAMP2 constructs mutated in

218 both cholesterol-binding sites could not be tested, as they were only poorly delivered to

219 lysosomes.

220

221 These experiments demonstrate a direct role for LAMP2 in cholesterol export from lysosomes,

222 and confirm that LAMP2’s ability to bind cholesterol correlates with its ability to support

223 cholesterol export from LAMP-deficient MEF cells. In addition, LAMP proteins bind tightly to

224 NPC proteins in vitro and in cells, and appear to facilitate cholesterol export from lysosomes.

225

226 LAMP proteins are the most highly abundant membrane glycoproteins of the lysosome, and their

227 lumenally oriented cholesterol binding sites represent a significant binding site for this important

228 sterol. We measured ~7 X 106 molecules per HeLa cell, representing 0.3mM binding sites in

10 229 lysosomes. A recent cellular mass spectrometry analysis (26) estimated LAMP proteins to be

230 present at 260,000 copies and NPC1 at 29,193 copies per HeLa cell. While the relative

231 abundance of these proteins matches the values we report here, their total level was 25 fold lower

232 in that study. It is possible that these transmembrane glycoproteins were under-represented in

233 due to their unusual resistance as proteins of the lysosome membrane, differences in cell

234 confluency and/or differences in HeLa cell lines employed.

235

236 Lysosomes have recently been shown to sense and signal amino acid availability to influence

237 lysosome biogenesis in relation to cellular need (27), and LAMP oligomerization has been

238 reported to correlate with chaperone mediated autophagy (11). Cholesterol levels may influence

239 LAMP protein conformation or interaction with other partners to signal the availability of

240 endocytosed cholesterol to influence autophagy and cellular metabolism. The ten fold higher

241 abundance of LAMP proteins compared with NPC1 protein in HeLa cells suggests that LAMP

242 proteins may do more than just facilitate NPC1 function in cholesterol export. Future

243 experiments will be needed to fully understand the roles played by these highly abundant

244 lysosomal membrane glycoproteins.

245

246 We have shown that LAMP2 binds tightly to the N-terminal domain of NPC1 and also binds

247 cholesterol with the same orientation as that domain. LAMP2 also aids in cholesterol export

248 from lysosomes. How might LAMP2’s cholesterol binding site contribute to cholesterol export?

249 Current models suggest that the soluble NPC2 protein binds cholesterol from the internal

250 membranes of lysosomes and delivers it to NPC1 at the limiting membrane of this compartment

251 (16). One possibility is that NPC2 can deliver cholesterol to both NPC1 and to LAMP2, which

11 252 is more abundant. This would help drive the cholesterol export process by moving cholesterol

253 from the accumulated, lumenal lipid stores to the lysosome’s limiting membrane. Because

254 LAMP2 and NPC1 N-terminal domains bind cholesterol in the same orientation, it makes sense

255 that NPC2 (which binds in opposite orientation, ref. 28) could transfer the cholesterol between

256 these two proteins. The recent crystal structure of NPC2 bound to the middle, lumenal domain

257 of NPC1 (29) supports a direct handoff between NPC1 and NPC2 (16). In future work, it will be

258 important to elucidate precisely how LAMP2 interacts with both NPC2 and NPC1 to facilitate

259 cholesterol export from lysosomes and how cholesterol binding contributes to LAMP2’s other

260 cellular roles.

261

262 Materials and Methods

263 Cholesterol, epicholesterol and sodium cholesteryl sulfate were from Sigma (St. Louis, MO); 24-

264 hydroxycholesterol (24-HC) was a gift from Rajat Rohatgi (Stanford University, Stanford, CA);

265 25-hydroxycholesterol and 7-ketocholesterol were from Steraloids (Newport, RI) or Avanti Polar

266 Lipids (Alabaster, AL); [1,2-3H]cholesterol (50 Ci/mmol) and 25-[26,27-3H] hydroxycholesterol

267 were from American Radiolabeled Chemicals (St. Louis, MO). Ni-NTA agarose was from

268 Qiagen (Valencia, CA); freestyle 293™ expression medium and Dulbecco’s modified Eagle’s

269 medium (DMEM) was from Life Technologies (Carlsbad, CA); lipoprotein deficient serum was

270 from KALEN Biomedical (Montgomery Village, Maryland). PierceTM Protein Concentrators

271 PES were from Thermo Fisher Scientific (Grand Island, NY); PD-10 desalting columns and Q-

272 Sepharose were from GE Healthcare Life Sciences (Pittsburgh, PA); pFastBac NPC1-N-terminal

273 domain plasmid, LAMP1-mGFP and MCOLN1-pEGFP C3 were from Addgene (Cambridge,

274 MA); pGEM-LAMP2 was from Sino Biological Inc; mouse anti-human LAMP1 and LAMP2

275 antibody culture supernatants were from Developmental Studies Hybridoma Bank (University of

276 Iowa, Iowa city, IA). Chicken anti-GFP antibody was from Aves Labs (Tigard, Oregon); IRDye

12 277 800CW donkey anti-chicken and IRDye 680RD donkey anti-mouse antibodies were from LI-

278 COR, Inc. (Lincoln, NE); anti-chicken-HRP conjugate was from Promega (Sunnyvale, CA); goat

279 anti-mouse-HRP conjugate was from BioRad (Hercules, CA); ECL Western Blotting Substrate

280 was from Thermo Scientific (Rockford, IL).

281 282 Buffers Buffer A: 50mM ammonium acetate, pH4.5, 150mM NaCl, 0.004% NP-40; buffer B: 283 50mM MES, pH5.5, 150mM NaCl, 0.004% NP40; buffer C: 50mM MES, pH6.5, 150mM NaCl, 284 0.004% NP-40; buffer D: 50mM HEPES, pH7.5, 150mM NaCl, 0.004% NP-40; buffer E: 25mM 285 Tris, pH7.4, 150mM NaCl; RIPA buffer: 50mM Tris, pH7.4, 150mM NaCl, 1% NP-40, 0.2% 286 deoxycholic acid, 0.1% SDS. 287 288 Plasmids cDNAs encoding full length, soluble human LAMP1(1-382), human LAMP2 (1-375) 289 and domain 1 of human LAMP2 (1-231) were PCR amplified from LAMP1-mGFP and pGEM- 290 LAMP2 respectively. The PCR products were inserted into pEGFP-N3 vector. The constructs 291 were assembled to have an unstructured GSTGSTGSTGA linker at the C terminus, followed by

292 a His10 tag and a FLAG tag. For LAMP2, another His10 tag was added downstream of the FLAG 293 tag for improved purification. LAMP2 domain 2 was prepared by deleting residues 39-219 from 294 the full length, soluble domain construct. FUGENE6 was used for transient transfection of HeLa 295 cells. Membrane anchored rescue constructs were stably expressed in LAMP1/2 deficient MEF 296 cells by lentivirus transduction and were comprised of full length LAMP2 bearing a C-terminal 297 GFP (LAMP2-GFP), or LAMP2-GFP Δ194-368 (encoding membrane anchored domain 1) or the 298 latter construct carrying point mutations. 299 300 Cell lines Authenticated HEK293F, HEK293T, and HeLa cells were from ATCC and used at 301 low passage; Sf9 cells were purchased from Thermo Fisher Scientific (Waltham, MA); Mouse 302 embryonic fibroblasts from LAMP1/LAMP2 double knockout mice (11,12) were the generous 303 gift of Dr. Paul Saftig (Christian-Albrechts-Universität Kiel, Germany). Mycoplasma 304 contamination was monitored by DAPI staining. 305

13 306 Cell culture All cells were cultured at 37°C and under 5% CO2 in Dulbecco’s modified Eagle’s 307 medium supplemented with 7.5% fetal bovine serum, 100U/ml penicillin and 100 μg/ml 308 streptomycin, unless indicated. HEK293F suspension cells were cultured at 37°C under 5% CO2 309 in Freestyle 293™ medium. In some experiments, cells were cultured in lipoprotein deficient 310 serum (5%). 311 312 Protein purification pFastBac NPC1-N-terminal domain plasmid was used to make virus for 313 infection of Sf9 insect cells. Seventy-two hours after infection, Sf9 cultures were spun down and 314 ammonium sulfate added to achieve 60% saturation. The resulting precipitate was re-suspended 315 buffer E and incubated with Ni-NTA resin overnight at 4°C. After washing with buffer E with 316 25mM imidazole, the protein was eluted with buffer E plus 250mM imidazole, and further 317 purified using Q-Sepharose. 318 319 HEK293F cells were transfected using 293fection according to the manufacturer. After 72h, 320 supernatants were collected after spinning 3000 rpm for 5min. To purify proteins for [3H] 321 cholesterol binding, supernatants were subjected to 90% ammonium sulfate precipitation. After 322 spinning at 13000 rpm for 30min, pellets were re-suspended in buffer E plus 25mM imidazole 323 and incubated with Ni-NTA resin overnight at 4°C, followed by washing with the same buffer. 324 Bound proteins were eluted with buffer E plus 250mM imidazole. Proteins were concentrated 325 and buffer exchanged into buffer C with PierceTM Protein Concentrators PES (10kD cut-off). 326 Proteins were either used immediately or stored at -80°C after snap freezing in liquid nitrogen. 327 For cholesterol extraction and thin layer chromatography, supernatants were adjusted to pH 7.4 328 and incubated with Ni-NTA resin overnight at 4°C; after washing with buffer E plus 25mM 329 imidazole, proteins were eluted with buffer E plus 250mM imidazole. Proteins were desalted into 330 PBS using a PD-10 column. 331 332 3H-cholesterol binding Each reaction was carried out in a final volume of 80–100 µl of buffer 333 A, B, C or D containing 0.1-1µg purified His-tagged protein and 10–400 nM 3H-cholesterol 334 diluted with 0.1-50µM cholesterol. For competition assays, reactions were in 80µl buffer C 335 (50mM MES, 150mM NaCl, 0.004% NP-40, pH6.5) with 0.1µg full length soluble LAMP2 336 protein and 50nM 3H-cholesterol, competition was started by adding vehicle (ethanol) or

14 337 different concentrations of competitors as indicated. After incubation overnight at 4°C, the 338 mixture was loaded onto a column packed with 30µl Ni-NTA agarose beads. After incubation for 339 10min, each column was washed with 5 ml of buffer C plus 10mM imidazole. The protein-bound 340 3H-cholesterol was eluted with 250mM imidazole-containing buffer C and quantified by 341 scintillation counting. For competition experiments, assays were carried out in the presence of 342 30µM unlabeled sterol. 343 344 Mass Spectrometry Samples were analyzed by LC/MS on an Agilent 1260 HPLC and Bruker 345 microTOF-Q II mass spectrometer. Full scan mass and product ion spectra were acquired in 346 positive ion mode, using a Phenomenex Kinetex C18 2.6u 2.1x100mm column, and an initial 347 condition of 30%, 0.1% formic acid in water/70% methanol. 348 349 Thin Layer Chromatography Full length soluble LAMP2, and domains 1 and 2 of LAMP2 350 were purified as described above. Extraction was performed by adding 3 sample volumes of 351 chloroform/methanol (2:1, v/v) to the samples. After repeating once more, extracts were pooled 352 and dried under nitrogen. The extracts were re-dissolved in 50-100µl chloroform/methanol (2:1, 353 v/v). Samples were spotted onto a Silica gel plate. The plate was developed with isopropanol 354 until the front reached 1cm above the loading position; after drying under airflow, the plate was 355 further developed using 2% methanol in chloroform until the front reached the top of the plate.

356 The plate was sprayed with 10% CuSO4 in 4% or 8% phosphoric acid and heated at 180°C to 357 visualize the samples. 358 359 Co-immunoprecipitation HEK293T cells expressing pEGFP-N1, pEGFP-N1-mNPC1 or 360 pEGFP-C3-MCOLN1 were harvested 24–48 h post-transfection and lysed in lysis buffer (50 mM 361 MES, pH 5.5, 150 mM NaCl and 0.1% digitonin) supplemented with protease inhibitors. After 362 30min on ice, lysates were spun at 15,000 g for 15 min, and protein concentrations of the 363 supernatants were measured. Equal amounts of extract protein were incubated with GFP-binding 364 protein–conjugated agarose for 2 h at 4°C. Immobilized proteins were washed 4 times with 1ml 365 lysis buffer, eluted with 2× SDS loading buffer, and subjected to BioRad Mini-PROTEIN TGX 366 4-20% gradient gels. After transfer to nitrocellulose membrane and antibody incubation, blots

15 367 were detected with ECL western blotting detection substrate or visualized using LI-COR 368 Odyssey Imaging System. 369 370 Microscale Thermophoresis (MST) 371 MST experiments were performed on a Monolith NT.115Pico instrument (Nanotemper

372 Technologies). Briefly, His6-NPC1 N-terminal domain, RNase B or NPC2 were labeled using 373 the RED-NHS (Amine Reactive) Protein Labeling Kit (Nanotemper Technologies). A constant

374 concentration of 6nM labeled protein was mixed with binding partners with a final buffer 375 condition of 50 mM MES, pH 5.5, containing 150 mM NaCl, 0.004% NP-40. Premium coated 376 capillaries contained 16 sequential, 2 fold serial dilutions. Analysis was at 40% laser power for 377 30 sec, followed by 5 sec cooling. Data were normalized to fraction of bound (0=unbound,

378 1=bound). The dissociation constant KD was obtained by plotting the normalized fluorescence

379 Fnorm against the logarithm of the different concentrations of the dilution series according to the 380 law of mass action. 381 382 Quantitation of intracellular LAMP2 and NPC1 protein 383 HeLa and HEK293T cells were grown to sub-confluence in DMEM supplemented with 7.5% 384 FBS. One 10cm dish of cells was washed 3 times with cold PBS, then lysed with 500µl RIPA 385 buffer with protease inhibitor cocktail (Sigma). After 30min on ice, the lysate was centrifuged at 386 13000 rpm for 15min at 4°C. The resulting supernatant was transferred to a new tube and 387 protein was measured by BCA assay. Lysates were resolved by SDS-PAGE, using different 388 amounts of purified human LAMP2 or NPC1 protein as standards. After transfer to 389 nitrocellulose, the blot was probed with anti-human LAMP2 or NPC1 antibody followed by 390 IRDye 800CW labeled anti mouse (for LAMP2) or rabbit (for NPC1) secondary antibody, and 391 visualized using a LI-COR Odyssey Imaging System and analyzed using ImageJ software. 392 Calculations were based on molecular weights of 45874 for LAMP2 and 142167 for NPC1 393 polypeptide chains, and neglected glycan contribution, which is not measured in the protein 394 assay employed. Purified, full length NPC1 protein was the gift of Dr. Xiaochun Li (Rockefeller 395 University) and was N-glycanase treated. 396

16 397 Other Methods Confocal immunofluorescence microscopy was carried out as described (32). 398 Cells grown on coverslips were fixed with 3.7% (vol/vol) paraformaldehyde for 15 min at room 399 temperature. LAMP1 staining was performed with sequential incubation of mouse anti-LAMP1 400 culture supernate and Alexa Fluor 594 goat anti-mouse antibody (1:1,000, Invitrogen), each for 1 401 h at room temperature. Coverslips were mounted using Mowiol and imaged using a Leica SP2 402 confocal microscope and Leica software with a 60×1.4 N.A. Plan Apochromat oil immersion 403 lens and a charge-coupled device camera (CoolSNAP HQ, Photometrics). Flow cytometry was 404 carried out on a FACScan Analyzer on gently trypsinized cells fixed as described above (32). 405 Structures were presented in drawings created using Chimera software (33).

406 Cholesterol Ester Formation LAMP1/LAMP2-deficient MEF cells were cultured in DMEM 407 medium with 5% (vol/vol) LPDS for two days and assayed (25, 32) with minor modification. 408 After 48 h, 100 µg/mL LDL, 50 µM lovastatin, and 50 µM sodium mevalonate were added for 409 5h. Cells were pulse labeled for 4 h with 0.1 mM sodium [1-14C]oleate (American Radiolabeled 410 Chemicals)–albumin complex. Cells were washed two times with 2 mL 50 mM Tris, 150 mM 411 NaCl, 2 mg/mL BSA, pH 7.4, followed by 2 mL 50 mM Tris, 150 mM NaCl, pH 7.4. Cells were 412 extracted and rinsed with hexane-isopropanol (3:2), pooled, and evaporated. After resuspending 413 each sample in 60 µl hexane, 4 µL of lipid standard containing 8 µg/mL triolein, 8 µg/mL oleic 414 acid, and 8 µg/mL cholesteryl oleate was added. Samples were spotted onto a silica gel 60 plastic 415 backed, thin layer chromatogram and developed in hexane. Cholesteryl oleate was identified 416 with iodine vapor, scraped from chromatograms, and radioactivity determined by scintillation 417 counting in 10 mL Biosafe II. 418 419 Statistical Analysis-- Minimum sample sizes were determined assuming 5% standard error 420 and > 95% confidence level. P values were determined using Graphpad Prism software and are 421 indicated in all figures according to convention: * P ≤ 0.05; ** P ≤ 0.01; *** P ≤ 0.001; **** P ≤ 422 0.0001. Error bars represent standard error of the mean. 423

424

17 425

426 References

427 1. Xu H, Ren D. 2015. Lysosomal physiology. Annu Rev Physiol 77:57-80. 428 2. Schwake M, Schröder B, Saftig P. 2013. Lysosomal membrane proteins and their central role in physiology. 429 Traffic 14:739-748. 430 3. Saftig P, Klumperman J. 2009. Lysosome biogenesis and lysosomal membrane proteins: trafficking meets 431 function. Nat Rev Mol Cell Biol 10:623-35. 432 4. Wilke S, Krausze J, Büssow K. 2012. Crystal structure of the conserved domain of the DC lysosomal associated 433 membrane protein: implications for the lysosomal glycocalyx. BMC Biol 10:62-77. 434 5. Kundra R, Kornfeld S. 1999. Asparagine-linked oligosaccharides protect Lamp-1 and Lamp-2 from intracellular 435 proteolysis. J Biol Chem 274:31039-31046. 436 6. Granger BL, Green SA, Gabel CA, Howe CL, Mellman I, Helenius A. 1990. Characterization and cloning of 437 lgp110, a lysosomal membrane glycoprotein from mouse and rat cells. J Biol Chem 265:12036-12043. 438 7. Tanaka Y et al. 2000. Accumulation of autophagic vacuoles and cardiomyopathy in LAMP-2-deficient mice. 439 Nature 406:902-906. 440 8. Huynh KK, Eskelinen EL, Scott CC, Malevanets A, Saftig P, Grinstein S. 2007. LAMP proteins are required for 441 fusion of lysosomes with phagosomes. EMBO J 26:313-324. 442 9. Cuervo AM, Dice JF. 1996. A receptor for the selective uptake and degradation of proteins by lysosomes. 443 Science 273:501-503. 444 10. Cuervo AM, Dice JF. 2000. Unique properties of lamp2a compared to other isoforms. J Cell Sci 113 Pt 445 24:4441-50. 446 11. Bandyopadhyay U, Kaushik S, Varticovski L, Cuervo AM. 2008. The chaperone-mediated autophagy receptor 447 organizes in dynamic protein complexes at the lysosomal membrane. Mol Cell Biol 28:5747-5763. 448 12. Eskelinen EL et al. 2004. Disturbed cholesterol traffic but normal proteolytic function in LAMP-1/LAMP-2 449 double-deficient fibroblasts. Mol Biol Cell 15:3132-45. 450 13. Schneede A et al. 2011. Role for LAMP-2 in endosomal cholesterol transport. J Cell Mol Med 15:280-295. 451 14. Hulce JJ, Cognetta AB, Niphakis MJ, Tully SE, Cravatt BF. 2013. Proteome-wide mapping of cholesterol- 452 interacting proteins in mammalian cells. Nat Methods 10:259-64. 453 15. Infante RE, Abi-Mosleh L, Radhakrsihnan A, Dale JD, Brown MS, Goldstein JL. Purified NPC1 protein: 1. 454 Binding of cholesterol and oxysterols to a 1278 amino acid membrane protein. J Biol Chem, 283, 1052-1063 455 (2008). 456 16. Kwon HJ et al. 2009. Structure of N-terminal domain of NPC1 reveals distinct subdomains for binding and 457 transfer of cholesterol. Cell 137, 1213-24 (2009). 458 17. A.I. Rosenbaum, F.R. Maxfield, Niemann-Pick type C disease: molecular mechanisms and potential therapeutic 459 approaches. J Neurochem 116, 789-795 (2011).

18 460 18. Abi-Mosleh L, Infante RE, Radhakrishnan A, Goldstein JL, Brown MS. 2009. Cyclodextrin overcomes 461 deficient lysosome-to-endoplasmic reticulum transport of cholesterol in Niemann-Pick type C cells. Proc Natl 462 Acad Sci USA 106:19316-21

463 19. Rosenbaum AI, Zhang G, Warren JD, Maxfield FR. 2010. Endocytosis of beta-cyclodextrins is responsible for 464 cholesterol reduction in Niemann-Pick type C mutant cells. Proc Natl Acad Sci USA 107:5477-82.

465 20. Ulloth JE, Almaguel FG, Padilla A, Bu L, Liu JW, De Leon M. 2007. Characterization of methyl-beta- 466 cyclodextrin toxicity in NGF-differentiated PC12 cell death. Neurotoxicology 28:613-21. 467 21. Lu F, Liang Q, Abi-Mosleh L, Das A, De Brabander JK, Goldstein JL, Brown MS. 2015. Identification of 468 NPC1 as the target of U18666A, an inhibitor of lysosomal cholesterol export and Ebola infection. Elife. 2015 469 Dec 8;4. pii: e12177. 470 22. Li X, Wang J, Coutavas E, Shi H, Hao Q, Blobel G. 2016. Structure of human Niemann-Pick C1 protein. Proc 471 Natl Acad Sci USA 113:8212-7. 472 23. Griffiths G, Back R, Marsh M. A quantitative analysis of the endocytic pathway in Baby Hamster Kidney 473 Cells. J. Cell Biol. 109, 2703-2720 (1989). 474 24. Quinn P, Griffiths G, Warren G. 1984. Density of Newly Synthesized Plasma Membrane proteins in 475 Intracellular membranes II. Biochemical Studies. J Cell Biol 98:2142-2147. 476 25. J.L. Goldstein, S.K. Basu SK, M.S. Brown, Receptor-mediated endocytosis of low-density lipoprotein in 477 cultured cells. Methods Enzymol 98, 241–260 (1983). 478 26. Itzhak DN, Tyanova S, Cox J, Borner GHH. 2016. Global, quantitative and dynamic mapping of protein 479 subcellular localization. eLife 2016;5:e16950. 480 27. Settembre C, Fraldi A, Medina DL, Ballabio A. Signals from the lysosome: a control centre for cellular 481 clearance and energy metabolism. Nat Rev Mol Cell Biol 14, 283-96 (2013). 482 28. Xu S, Benoff B, Liou HL, Lobel P, Stock AM. 2007. Structural basis of sterol binding by NPC2, a lysosomal 483 protein deficient in Niemann-Pick type C2 disease. J Biol Chem 282:23525-31. 484 29. Li X, Saha P, Li J, Blobel G and Pfeffer SR. Clues to the mechanism of cholesterol transfer from the structure 485 of NPC1 middle lumenal domain bound to NPC2. Proc Natl Acad Sci USA in press (2016)

486 30. Vergarajauregui S1, Martina JA, Puertollano R. 2011. LAPTMs regulate lysosomal function and interact with 487 mucolipin 1: new clues for understanding mucolipidosis type IV. J Cell Sci 124(Pt 3):459-68.

488 31. Gong X, Qian H, Zhou X, Wu J, Wan T, Cao P, Huang W, Zhao X, Wang X, Wang P, Shi Y, Gao GF, Zhou Q, 489 Yan N. 2016. Structural Insights into the Niemann-Pick C1 (NPC1)-Mediated Cholesterol Transfer and Ebola 490 Infection. Cell 165:1467-78. 491 32. Li J, Deffieu MS, Lee PL, Saha P, Pfeffer SR. 2015. Glycosylation inhibition reduces cholesterol accumulation 492 in NPC1 protein-deficient cells. Proc Natl Acad Sci U S A. 112:14876-81. 493 33. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, Ferrin TE. 2004. UCSF Chimera- 494 a visualization system for exploratory research and analysis. J Comput Chem 25:1605-12.

495

19 496 Acknowledgments: This research was funded by a grant from the Ara Parseghian Medical

497 Research Foundation and NIH DK37332 to SRP. We are grateful to Drs. Christopher Wassif

498 and Forbes Denny Porter (NIH) for independent confirmation of the mass spectrometry results

499 and Dr. Paul Saftig for providing MEF cells lacking LAMP1/LAMP2.

500

501 Figure 1. Cholesterol binding to LAMP proteins. A. Coomassie-stained SDS-PAGE of

502 purified, secreted human LAMP1, LAMP2 or LAMP domains from LAMP2. B,C. 3H-

503 cholesterol or 3H-25 hydroxycholesterol binding to soluble LAMP2 (full length protein). Also

504 shown in C is binding in the presence of 50µM cold hydroxycholesterol. D,E, 3H-cholesterol

505 binding to indicated, soluble proteins compared with the soluble NPC1 N-terminal domain. F,

506 Cholesterol binding to LAMP2 domains 1 and 2 compared with full length, soluble LAMP2

507 protein. P values were determined relative to the full length soluble LAMP2 protein. G. Sterol

508 competition (30µM) for 3H-cholesterol binding to soluble LAMP2. P values were determined

509 relative to no cold addition. H, I. pH dependence and kinetics of 3H-cholesterol binding to

510 LAMP2. In B, C, E, H and I, a representative experiment is shown; C and E show the average of

511 duplicates. In C, the background counts in reactions containing the control protein, GFP-binding

512 protein, were subtracted; in D and E, the control was TIP47 protein. D, F, and G show the

513 combined results of two experiments in duplicate. Numbers at right (A) indicate mass in kD for

514 this and all subsequent figures. Reactions contained D, E, G, H, I, 500nM total cholesterol; F,

515 5µM total cholesterol; B and C were carried out using increasing concentrations of the indicated

516 sterol.

517

518 Figure 1-figure supplement 1. Diagram of constructs used.

20 519

520 Figure 2. Cholesterol binding to LAMP2 is competed by 25-hydroxycholesterol (B), cholesterol

521 (C), 7-ketocholesterol (D) but not cholesterol sulfate (A) or epicholesterol (C). The structures

522 above indicate the regions of the sterol that differ from cholesterol. In C, the background

523 obtained in reactions containing GFP was subtracted. All panels used 50nM 3H-cholesterol and

524 the indicated amounts of competitors.

525

526 Figure 3. Mass spectrometry identification of small molecules released from LAMP2 after

527 chloroform:methanol (2:1) extraction. A, masses from cholesterol standard; B, masses of

528 LAMP2-bound material; C, Copper sulfate/phosphoric acid detection of indicated markers after

529 thin layer chromatography compared with material eluted from soluble LAMP2 (50µg). Shown

530 are the results of a representative experiment carried out twice.

531

532 Figure 4. LAMP2 interacts with NPC1 and NPC2 proteins. A, Anti-GFP immunoprecipitation

533 from HEK293T cells grown in FBS containing medium expressing GFP or mouse NPC1-GFP.

534 The blot was developed with ECL. Upper panel, anti-GFP immunoblot (100% elution); lower

535 panel, anti-LAMP2 immunoblot to detect endogenous, full length protein (100% elution). B, C,

536 co-immunoprecipitation of LAMP2 and NPC1-GFP in cells grown in FBS, LPDS (5%) or LPDS

537 + cyclodextrin (1mM) for 24 hours. Left panels, 1% inputs; right panels, 50% elutions. Error

538 bars represent SEM for two combined experiments carried out in duplicate; P value is from

539 comparison with LPDS by two-tailed Student’s t-test. B shows duplicate reactions to document

540 reproducibility. D, Anti-GFP immunoprecipitation of HEK293T cells grown in FBS expressing

541 GFP-MCOLN1 or mouse NPC1-GFP. GFP-MCOLN1 occurs as a ~90kD form, a proteolytically

21 542 processed form, and as higher oligomers (30). Left panels, inputs (2%); right panels, total

543 elution carried out in duplicate. GFP proteins are green; LAMP2 is presented in red. Shown is a

544 representative experiment carried out twice in duplicate.

545

546 Figure 5. LAMP2 binds NPC1 and NPC2 proteins. A, B, Structures of NPC1 (31; pdb 3jD8)

547 and NPC2 (28; pdb 2hka) proteins; C,E, Microscale thermophoresis (E) or fluorescence (C)

548 obtained with mixtures of soluble, AF647 labeled-NPC1 N-terminal domain or AF647-RNAase

549 B with increasing concentrations of soluble LAMP2 in 1µM cholesterol. D,F, Microscale

550 thermophoresis of AF647 labeled-NPC2 with increasing concentrations of soluble LAMP2 or

551 RNAase B in 1µM cholesterol sulfate. For C-F, a representative experiment carried out at least

552 twice is shown.

553

554 Figure 5 -- figure supplement 1. Quantitation of NPC1 (A) and LAMP2 (B) proteins in HeLa

555 and HEK293T cells. The values at the top of each gel represent the amount of purified protein or

556 total cell extract analyzed, determined by BCA protein assay. The immunoblot was developed

557 using rabbit anti-NPC1 or mouse anti-LAMP2 antibodies followed by detection with

558 IRDye800CW goat anti-mouse or donkey anti-rabbit antibodies. Shown is an example of an

559 experiment carried out in duplicate; subconfluent cultures were analyzed. Purified NPC1 is

560 slightly smaller than that in the extract because it was deglycosylated (see Methods).

561

562 Figure 6. A, predicted structure model of LAMP2 domain 1; residues I111 and V114 are

563 highlighted in red. B, Relative 3H-cholesterol binding to soluble LAMP 2 domain 1 or LAMP2

564 domain 1-I111A/V114A. Shown is combined data from 5 independent experiments carried out in

22 565 duplicate in the presence of 50nM 3H-cholesterol. Inset, SDS-PAGE analysis of wild type (left)

566 and domain 1-I111A/V114A (right) proteins analyzed. P value was determined by two-tailed

567 Student’s t-test. C, flow cytometry analysis of mean fluorescence of GFP rescue constructs in

568 lentivirus-tranduced cells (>20,000 cells analyzed). D , Cholesteryl oleate synthesis in MEF

569 cells lacking LAMP1 and LAMP2 after rescue with either full length, membrane anchored

570 LAMP2, membrane anchored LAMP2 domain 1, or membrane anchored LAMP2 domain 1-

571 I111A/V114A. C-terminally GFP-tagged, rescue proteins were stably expressed using lentivirus

572 transduction; shown is the combined result of 2 independent experiments, normalized for the

573 amount of mature protein in each sample (Fig. 6-Supp. Fig. 2) relative to the amount of rescue

574 seen with full length LAMP2 protein. P-values are in relation to full length for domain 1, or to

575 domain 1 for the mutant protein, and were determined by one way ANOVA. E, confocal

576 microscopic analysis of GFP rescue construct localization (green) and endogenous LAMP1

577 protein (red) in transiently transfected HeLa cells; bars represent 20µm.

578

579 Figure 6 -- figure supplement 1. Immunoblot analysis of LAMP2 constructs from lentivirus

580 transduced LAMP1/LAMP2-knock out MEF cells. A, total cell extract from cells expressing

581 membrane anchored, GFP-LAMP2 domain 1 or GFP-LAMP2 domain 1-I111A/V114A; B,

582 migration of the constructs indicated, before or after 4 hours cycloheximide treatment (50

583 µg/ml). Panel A was developed as in Fig. 4B; panel B was developed as in Fig. 4A.

23